CSE 403Lecture 14

Integration Testing

Reading:

The Art of Unit Testing, Ch. 1, 3, 4-5 (Osherove)Code Complete, Ch. 29 (McConnell)

slides created by Marty Stepphttp://www.cs.washington.edu/403/

2

Integration

• integration: Combining 2 or more software units– often a subset of the overall project (!= system testing)

• Why do software engineers care about integration?

– new problems will inevitably surface•many systems now together that have never been before

– if done poorly, all problems present themselves at once•hard to diagnose, debug, fix

– cascade of interdependencies•cannot find and solve problems one-at-a-time

3

Phased integration

• phased ("big-bang") integration:– design, code, test, debug each class/unit/subsystem

separately– combine them all– pray

4

Incremental integration

• incremental integration:– develop a functional "skeleton" system (i.e. ZFR)– design, code, test, debug a small new piece– integrate this piece with the skeleton

•test/debug it before adding any other pieces

5

Benefits of incremental

• Benefits:– Errors easier to isolate, find, fix

•reduces developer bug-fixing load

– System is always in a (relatively) working state•good for customer relations, developer morale

• Drawbacks:– May need to create "stub" versions of some features that

have not yet been integrated

6

Top-down integration

• top-down integration: Start with outer UI layers and work inward– must write (lots of) stub lower layers for UI to interact

with– allows postponing tough design/debugging decisions

(bad?)

7

Bottom-up integration

• bottom-up integration: Start with low-level data/logic layers and work outward– must write test drivers to run these layers– won't discover high-level / UI design flaws until late

8

"Sandwich" integration

• "sandwich" integration: Connect top-level UI with crucial bottom-level classes– add middle layers later as needed– more practical than top-down or bottom-up?

9

Daily builds

• daily build: Compile working executable on a daily basis– allows you to test the quality of your integration so far– helps morale; product "works every day"; visible progress– best done automated or through an easy script– quickly catches/exposes any bug that breaks the build

• smoke test: A quick set of tests run on the daily build.– NOT exhaustive; just sees whether code "smokes" (breaks)– used (along with compilation) to make sure daily build runs

• continuous integration:Adding new units immediately as they are written.

10

Integration testing

• integration testing: Verifying software quality by testing two or more dependent software modules as a group.

• challenges:– Combined units can fail

in more places and in morecomplicated ways.

– How to test a partial systemwhere not all parts exist?

– How to "rig" the behavior ofunit A so as to produce agiven behavior from unit B?

11

Stubs

• stub: A controllable replacement for an existing software unit to which your code under test has a dependency.

– useful for simulating difficult-to-control elements:•network / internet•database•time/date-sensitive code•files•threads•memory

– also useful when dealing with brittle legacy code/systems

12

Create a stub, step 1

• Identify the external dependency.– This is either a resource or a class/object.– If it isn't an object, wrap it up into one.

•(Suppose that Class A depends on troublesome Class B.)

13

Create a stub, step 2

• Extract the core functionality of the object into an interface.– Create an InterfaceB based on B– Change all of A's code to work with type InterfaceB, not B

14

Create a stub, step 3

• Write a second "stub" class that also implements the interface,but returns pre-determined fake data.– Now A's dependency on B is dodged and can be tested

easily.– Can focus on how well A integrates with B's external

behavior.

15

Injecting a stub

• seams: Places to inject the stub so Class A will talk to it.

– at construction (not ideal)

A aardvark = new A(new StubB());

– through a getter/setter method (better)

A apple = new A(...);aardvark.setResource(new StubB());

– just before usage, as a parameter (also better)

aardvark.methodThatUsesB(new StubB());

•You should not have to change A's code everywhere (beyond using your interface) in order to use your Stub B. (a "testable design")

16

"Mock" objects

• mock object: A fake object that decides whether a unit test has passed or failed by watching interactions between objects.

– useful for interaction testing (as opposed to state testing)

17

Stubs vs. mocks– A stub gives out data that goes to

the object/class under test.– The unit test directly asserts against

class under test, to make sure it givesthe right result when fed this data.

– A mock waits to be called bythe class under test (A).•Maybe it has several methods

it expects that A should call.

– It makes sure that it was contactedin exactly the right way.•If A interacts with B the way it should, the test passes.

18

Mock object frameworks• Stubs are often best created by hand/IDE.

Mocks are tedious to create manually.

• Mock object frameworks help with the process.– android-mock, EasyMock, jMock (Java)– FlexMock / Mocha (Ruby)– SimpleTest / PHPUnit (PHP)– ...

• Frameworks provide the following:– auto-generation of mock objects that implement a given

interface– logging of what calls are performed on the mock objects– methods/primitives for declaring and asserting your expectations

19

A jMock mock objectimport org.jmock.integration.junit4.*; // Assumes that we are testingimport org.jmock.*; // class A's calls on B.

@RunWith(JMock.class)public class ClassATest { private Mockery mockery = new JUnit4Mockery(); // initialize jMock

@Test public void testACallsBProperly1() { // create mock object to mock InterfaceB final InterfaceB mockB = mockery.mock(InterfaceB.class);

// construct object from class under test; attach to mock A aardvark = new A(...); aardvark.setResource(mockB);

// declare expectations for how mock should be used mockery.checking(new Expectations() {{ oneOf(mockB).method1("an expected parameter"); will(returnValue(0.0)); oneOf(mockB).method2(); }});

// execute code A under test; should lead to calls on mockB aardvark.methodThatUsesB();

// assert that A behaved as expected mockery.assertIsSatisfied(); }}

20

jMock API

• jMock has a strange API based on "Hamcrest" testing syntax.

• Specifying objects and calls:– oneOf(mock), exactly(count).of(mock), – atLeast(count).of(mock), atMost(count).of(mock), – between(min, max).of(mock)– allowing(mock), never(mock)

•The above accept a mock object and return a descriptor that you can call methods on, as a way of saying that you demand that those methods be called by the class under test.

– atLeast(3).of(mockB).method1();•"I expect that method1 will be called on mockB 3 times here."

21

Expected actions•.will(action)

– actions: returnValue(v), throwException(e)

• values:– equal(value), same(value), any(type), aNull(type), aNonNull(type), not(value), anyOf(value1, ..,valueN)

– oneOf(mockB).method1();

will(returnValue(anyOf(1, 4, -3)));

•"I expect that method1 will be called on mockB once here, and that it will return either 1, 4, or -3."

22

Using stubs/mocks together

• Suppose a log analyzer reads from a web service.If the web fails to log an error, the analyzer must send email.– How to test to ensure that this behavior is occurring?

• Set up a stub for the web service that intentionally fails.

• Set up a mock for the email service that checks to see whether the analyzer contacts it to send an email message.